Re: [PATCH] drivers/base/memory: make memory block get/put explicit

[David Hildenbrand (Arm)]

On 5/12/26 05:11, Muchun Song wrote:
> 
> 
>> On May 11, 2026, at 21:23, Muchun Song <muchun.song@xxxxxxxxx> wrote:
>>
>>
>>
>>>
>>> On 5/11/26 13:18, Muchun Song wrote:
>>>
>>> Better mention some of the other changes here, like removing find_memory_block().
>>
>> Will do.
>>
>>>
>>> [...]
>>>
>>>
>>> While at it, please drop the "extern".
>>
>> OK.
>>
>>>
>>>
>>> Would guards come in handy here?
>>
>> You mean to introduce something like:
>>
>> scoped_guard(memory_block, id) {
>> }
>>
>> Right? If yes, I will give it a try.
> 
> Hi David,

Hi,

> 
> Did I get that right?

I assume so, but it's indeed rather ugly. ... in particular the CLASS and
scoped_class() thingies are pretty intuitive.

I thought we could use guard()/scoped_guard(), but reading the details, it's
mostly for locks only.

There is only one users of scoped_class in the tree (overlayfs).

Maybe, if we would do this properly, we would actually provide our own wrappers,
like

	MEMORY_BLOCK
	scoped_memory_block

not even providing "memory_block_get", as that would be implicit.

Like

	MEMORY_BLOCK(mem)(block_id); // the second () is confusing
	scoped_memory_block(mem, block_id) {

	}


But that requires more thought, and I don't really know what the best practice
is there ...

So thanks for trying, but let's leave it as is for now.

-- 
Cheers,

David